The situation
Gusto runs payroll, benefits, and HR for hundreds of thousands of small businesses — a domain where AI has to be right, not just impressive. The company wanted agents in its internal tools and in customer-facing products like Gus, but every team building its own meant duplicated effort and inconsistent quality. What it needed was one platform underneath all of them.
What I did
Embedded as lead of the Agent Framework team, I set both the technical and the product direction for the AI platform behind Gusto's internal tools and customer-facing products. Same playbook as Filevine, bigger org: demos first, then land what ships with each team.
The other half of the job was making every team fast. I drove the AI primitives Gusto's agents are built from, worked out how to detect which use cases were actually worth an agent, and wrote the internal AI Wiki so teams could go from idea to shipped agent without me in the loop.
Platform work is deliberately unglamorous, and that's the point: the primitives are where an org becomes AI native or doesn't.
Where it stands
The Agent Framework powers Gus and Gusto's internal agents today, and teams across the org build on it rather than around it. Nobody mandated that. The working software made the case.
“The platform succeeded the day teams started shipping agents without asking me first.”